Three‐dimensional gradient and spin‐echo readout for time‐encoded pseudo‐continuous arterial spin labeling: Influence of segmentation factor and flow compensation

Abstract

Purpose
To monitor the complete passage of the labeled blood through the vascular tree into tissue and improve the quantification of ASL maps, we evaluated the effect of 3D gradient and spin‐echo (GRASE) readout segments on temporal SNR (tSNR) and image blurriness for time‐encoded pseudo‐continuous arterial spin labeling and the effect of flow‐compensation gradients on the presence of intravascular signal.

Methods
Fifteen volunteers were scanned using time‐encoded pCASL with 2D EPI and single‐segment, two‐segments, and three‐segments 3D‐GRASE readouts with first‐order flow compensation (FC) gradients. Two‐segments 3D‐GRASE scans were acquired with 25%, 50%, 75%, and 100% of full first‐order FC. Temporal SNR was assessed, and cerebral blood flow and arterial blood volume were quantified for all readout strategies.

Results
For single‐segment 3D GRASE, tSNR was comparable to 2D EPI for perfusion signal but worse for the arterial signal. Two‐segments and three‐segments 3D GRASE resulted in higher tSNR than 2D EPI for perfusion and arterial signal. The arterial signal was not well visualized for 3D‐GRASE data without FC. Visualization of the intravascular signal at postlabeling delays of 660 ms and 1060 ms was restored with FC. Adequate visualization of the intravascular signal was achieved from 75% of FC gradient strength at a postlabeling delay of 660 ms. For a postlabeling delay of 1060 ms, full‐FC gradients were the best option to depict intravascular signal.

Conclusion
Segmented GRASE provided higher effective tSNR compared with 2D‐EPI and single‐segment GRASE. Flow compensation with GRASE readout should be carefully controlled when applying for time‐encoded pCASL to visualize intravascular signal.

Full text

1454
|
Magn Reson Med. 2021;86:1454–1462.
wileyonlinelibrary.com/journal/mrm
Received: 22 December 2020
|
Revised: 14 March 2021
|
Accepted: 25 March 2021
DOI: 10.1002/mrm.28807
TECHNICAL NOTE
Three- dimensional gradient and spin- echo readout for time-
encoded pseudo- continuous arterial spin labeling: Influence of
segmentation factor and flow compensation
Andre M.Paschoal1,2,3,4
|
Renata F.Leoni2
|
Bruno F.Pastorello3
|
Matthias J. P.van Osch4
1Medical School of Ribeirao Preto, University of Sao Paulo, Ribeirao Preto, SP, Brazil
2InBrain Lab, Department of Physics - FFCLRP, University of Sao Paulo, Ribeirao Preto, SP, Brazil
3LIM44 - Instituto e Departamento de Radiologia, Faculdade de Medicina - Universidade de São Paulo, São Paulo, SP, Brazil
4C.J. Gorter Center for High Field MRI, Department of Radiology, Leiden University Medical Center, Leiden, The Netherlands
This is an open access article under the terms of the Creative Commons Attribution- NonCommercial License, which permits use, distribution and reproduction in any medium,
provided the original work is properly cited and is not used for commercial purposes.
© 2021 The Authors. Magnetic Resonance in Medicine published by Wiley Periodicals LLC on behalf of International Society for Magnetic Resonance in Medicine
Correspondence
Andre M. Paschoal, LIM44, Instituto e
Departamento de Radiologia, Faculdade de
Medicina, Universidade de São Paulo, São
Paulo, SP 05403- 010, Brazil.
Email: andre.paschoal@usp.br
Twitter: @pasch oal_am
Funding information
Coordenação de Aperfeiçoamento
de Pessoal de Nível Superior - Brasil
(CAPES) - PDSE, Grant/Award Number:
88881.188976/2018- 01; Conselho
Nacional de Desenvolvimento Científico
e Tecnológico (CNPq), Grant/Award
Number: 140110/2016- 0; Research program
in Applied and Engineering Sciences
- Netherlands Organization for Scientific
Research (NWO), Grant/Award Number:
VIC: 016.160.351
Purpose: To monitor the complete passage of the labeled blood through the vascular
tree into tissue and improve the quantification of ASL maps, we evaluated the effect
of 3D gradient and spin- echo (GRASE) readout segments on temporal SNR (tSNR)
and image blurriness for time- encoded pseudo- continuous arterial spin labeling and
the effect of flow- compensation gradients on the presence of intravascular signal.
Methods: Fifteen volunteers were scanned using time- encoded pCASL with 2D EPI
and single- segment, two- segments, and three- segments 3D- GRASE readouts with
first- order flow compensation (FC) gradients. Two- segments 3D- GRASE scans
were acquired with 25%, 50%, 75%, and 100% of full first- order FC. Temporal SNR
was assessed, and cerebral blood flow and arterial blood volume were quantified for
all readout strategies.
Results: For single- segment 3D GRASE, tSNR was comparable to 2D EPI for perfu-
sion signal but worse for the arterial signal. Two- segments and three- segments 3D
GRASE resulted in higher tSNR than 2D EPI for perfusion and arterial signal. The
arterial signal was not well visualized for 3D- GRASE data without FC. Visualization
of the intravascular signal at postlabeling delays of 660 ms and 1060 ms was restored
with FC. Adequate visualization of the intravascular signal was achieved from 75%
of FC gradient strength at a postlabeling delay of 660 ms. For a postlabeling delay of
1060 ms, full- FC gradients were the best option to depict intravascular signal.
Conclusion: Segmented GRASE provided higher effective tSNR compared with
2D- EPI and single- segment GRASE. Flow compensation with GRASE readout
should be carefully controlled when applying for time- encoded pCASL to visualize
intravascular signal.

|
1455
PASCHOAL et AL.
KEYWORDS
3D GRASE, flow compensation, time- encoded pCASL, tSNR
1
|
INTRODUCTION
Arterial spin labeling (ASL) is a noninvasive perfusion-
weighted imaging technique that exploits arterial blood as an
endogenous tracer. A bolus of arterial blood is magnetically la-
beled through inversion in a slab proximal to the imaging plane,
from which it will flow through the arterial tree until crossing
the blood- brain barrier and reaching the brain- tissue compart-
ment. After labeling the blood through the application of RF
pulses has stopped, the image is acquired after a time interval
called postlabeling delay (PLD). A single PLD is considered
adequate to measure cerebral blood flow (CBF).1 However,
when the subject suffers from prolonged arterial transit time
(ATT), CBF will be underestimated, which can be corrected by
acquiring a multiple- timepoint ASL (multi- PLD ASL).
Moreover, multi- PLD ASL allows monitoring the labeled
arterial blood within the intravascular space at short PLDs,
until finally perfusing the brain tissue at the later PLDs. By
using, for example, Bayesian inference under the appropriate
kinetic model, all PLDs are used to estimate quantitative ASL
maps, which include the CBF, arterial blood volume (aBV),
and ATT. In such a model, all PLDs acquired are important
for the final CBF maps to be corrected for arrival- time arti-
facts and intravascular signal.
In 2015, a consensus paper set the recommended imple-
mentation of ASL for clinical applications,1 to help with stan-
dardization and dissemination of the technique. In summary, a
segmented 3D gradient and spin- echo (GRASE) readout2- 4 was
recommended in combination with pseudo- continuous arterial
spin labeling (pCASL), as it can be optimally combined with
background suppression using a single excitation pulse per TR,
resulting in higher SNR compared with 2D EPI.5,6 However,
choosing the number of segments for 3D GRASE requires fine-
tuning to balance SNR, blurring in z- direction, and vulnerabil-
ity to motion. Segmented 3D sequences fill less k- space lines
per excitation pulse, thereby reducing signal losses due to T2
*
and consequently providing less blurring at boundaries between
tissues, especially where a significant signal difference exists
(eg, between CSF and gray matter). Therefore, segmented ac-
quisitions are less suited in situations in which many different
conditions need to be measured, such as in multi- postlabeling
delays (PLD) ASL. Although single- segment 3D GRASE pro-
vides a more time- efficient acquisition, it results in significant
blurring in z- direction, so it has been scarcely used.
There are three main approaches for multi- PLD ASL.
First, traditional pCASL scans can be repeated with dif-
ferent PLDs and labeling duration. Recently, Woods et al7
proposed optimal settings for such an acquisition. Second,
a Look- Locker readout can be used as, for example, in the
inflow turbo- sampling EPI– flow- sensitive alternating in-
version recovery and QUASAR (Modus QA, London, ON,
Canada) sequences.8,9 This approach has the advantage that
all PLDs are acquired simultaneously, albeit with lower SNR
resulting in similar total scan times. Finally, Hadamard-
encoded pCASL or time- encoded pCASL (te- pCASL) was
demonstrated, which achieves multiple PLDs by the tem-
poral encoding of the labeled blood following a Hadamard
matrix.10- 13 When designing a 3D readout sequence for multi-
PLD acquisitions, the criteria are different compared with
single- PLD perfusion imaging. Special attention is needed
when considering the use of flow compensation (FC) gradi-
ents that compensate for first- order dephasing effects.14 Use
of FC can increase TE and readout times, but its absence will
crush vascular ASL signal, which is an essential part of the
information conveyed in early PLDs by showing the passage
of label through the arterial system.
The appropriate readout setup and use of FC gradi-
ents provide the promise of images with higher temporal
SNR (tSNR) and improved delineation of the angiographic
phase, respectively, as an input for the quantitative model,
which would lead to more accurate CBF, aBV, and ATT
maps. Therefore, we aimed to evaluate the effect of the
number of segments of 3D- GRASE readout on tSNR, com-
paring its performance with the 2D- EPI readout, and ana-
lyze the effect of FC gradients in delineating the presence
of vascular signal.
2
|
METHODS
2.1
|
Hadamard- encoded pCASL
Time- encoded pCASL is a time- efficient strategy to acquire
multi- PLD pCASL, providing the opportunity to monitor
the dynamic inflow of blood into brain tissue, improving
CBF quantification, and allowing the measurements of ATT
and aBV. In a time- encoded acquisition, the labeling mod-
ule is divided into small subboli, encoded according to a
Hadamard matrix to be either control or label. In this study, a
Hadamard- 8 matrix was chosen for the labeling module, with
a total labeling duration of 3500 ms split into seven blocks of
1800, 800, 400, 2 × 150, and 2 × 100 ms (ie, the free- lunch
approach).13 After the labeling module, there was a PLD of
160 ms until starting the readout module (Figure 1B). Two
background suppression pulses were applied at 1831 ms and
3135 ms.

1456
|
PASCHOAL et AL.
2.2
|
Three- dimensional GRASE readout:
Number of segments
The 3D- GRASE readout combines the time efficiency of
EPI with higher SNR of 3D imaging by exploiting the rapid
imaging with refocused echoes readout15 technique in the
z- direction (Figure 1C). When an entire volume is acquired
using only a single excitation pulse, the approach is called
single- segment 3D GRASE. When more than one segment
is required to fill the 3D k- space, the approach is referred
to as multisegmented or segmented 3D GRASE. Because T2
signal decreases over the readout train result in blurring on
the kz- axis, GRASE segmentation is recommended. In this
study, we evaluated how the number of segments in 3D-
GRASE readout influences the te- pCASL acquisition perfor-
mance concerning time efficiency, blurring in the z- direction,
and SNR. Moreover, the results were compared with a 2D
multislice EPI readout.
2.3
|
Three- dimensional GRASE
readout: FC
The purpose of using FC gradients is to correct for flow-
induced dephasing. It can be achieved using additional bi-
polar gradient lobes before the signal readout (Figure 1D).
As the area under the positive and negative lobes are the same,
the added mean gradient is zero, but the timing of each gra-
dient lobe is chosen, so that the spins flowing with constant
velocity end up with zero phase accumulation over the total
readout (ie, no dephasing). For first- order FC, the dephas-
ing is only circumvented for constant velocity (ie, accelera-
tion [deceleration] of blood can still lead to dephasing). Due
to the added gradients, flow compensation will increase TE
value and lengthen the total readout time. Nonetheless, espe-
cially when using a 3D readout, visualization of intravascular
signal of dynamic ASL signal can be considered critical, es-
pecially during the inflow phase. Therefore, we analyzed the
FIGURE 1 A, Pulse sequence diagram of 3D time- encoded pseudo- continuous arterial spin labeling (te- pCASL). B, Eight Hadamard
matrix for pCASL labeling encoding, followed by the even postlabeling delay (PLD) intervals (in which the background suppression pulses
are applied) until reaching the readout module for acquisitions 1 to 8. C, Three- dimensional gradient and spin- echo (GRASE) readout module.
When the image is acquired with a single segment, the whole- brain acquisition is achieved with only one 90º RF excitation pulse, whereas for
two- segments and three- segments images, the readout module is repeated two and three times, respectively, (two and three 90º RF excitation
pulses) to acquire a whole- brain volume. D, When flow compensation is desired, bipolar gradient lobes are applied in the readout direction just
before the acquisition

|
1457
PASCHOAL et AL.
effect of FC gradients scaling for two- segments 3D- GRASE
readout on the resulting images, for both the angiographic
and the perfusion phase of multi- PLD time- encoded ASL.
2.4
|
Magnetic resonance imaging
experiments details
Fifteen volunteers (8 females, age 26.3 ± 5.2 years)
were scanned on a 3T Philips MRI scanner (Amsterdam,
Netherlands), from which 10 were scanned for the readout
segmentation experiment and five for the FC experiment. All
volunteers agreed to participate after signing informed con-
sent, and the study was in agreement with local internal review
board policies. We used a time- encoded Hadamard- 8 pCASL
labeling sequence with single- segment, two- segments, and
three- segments 3D GRASE and a multislice 2D- EPI readout.
Table 1 lists the acquisition parameters.
For the FC experiment, two- segments 3D- GRASE scans
were acquired with the FC gradient strengths of 0%, 25%,
50%, 75%, and 100% of full first- order strength to analyze the
effect of FC on the visualization of the vascular ASL signal.
2.5
|
Data analysis
Hadamard matrix decoding for ASL subtraction was per-
formed in MATLAB (MathWorks, Natick, MA) to obtain
multi- PLD ASL maps (note that labeling duration is also
different among the multi- PLD maps). Further processing of
ASL data and quantification of CBF and aBV were performed
using the BASIL toolkit (Oxford Center for Functional MRI
of the Brain’s software library [FSL]),16,17 considering all
PLDs acquired. Scans were corrected for motion using the
MCFLIRT tool (FSL), and the structural images were seg-
mented and transformed into the ASL space to create whole-
brain, gray- matter, and white- matter masks using FAST and
FLIRT tools (FSL).
Perfusion images were deblurred using the oxasl- deblur
toolkit of the Quantitative Biomedical Inference Group, Oxford
Institute of Biomedical Engineering (https://oxasl.readt he-
docs.io/en/lates t/downl oad.html), which uses a fast Fourier
transform filter in the frequency space to reduce the blurring
in z- direction for 3D- GRASE data. Furthermore, both blurred
and deblurred ASL perfusion maps were compared with the
gray- matter region of interest of the T1- weighted structural
images transformed to the ASL space (reference image). The
blurring estimation in z- direction was done in MATLAB using
a method that compares the desired image with a reference
image and outputs the blurring of the input image.18
Postprocessing and calculation of tSNR was done in
MATLAB according to:
where S is the mean signal of a voxel over the different aver-
ages, and
𝜎
is the SD. The perfusion tSNR was voxel- wise cal-
culated within the region of interest of the gray- matter mask
(1)
tSNR
=S
𝜎
∗
√
Number of averages in 6 min
Total averages acquired
,
TABLE 1 Acquisition parameters
Scan parameter 2D EPI
3D GRASE Single
segment
3D GRASE Two
segments
3D GRASE
Three segments
FOV (mm) 240 × 240 240 × 240 × 102 240 × 240 × 102 240 × 240 × 102
Voxel size (mm) 3.75 × 3.75 × 6 3.75 × 3.75 × 6 3.75 × 3.75 × 6 3.75 × 3.75 × 6
Acquisition matrix 64 × 64 × 17 64 × 64 × 31 64 × 64 × 31 64 × 64 × 31
Reconstructed slices 17 17 17 17
Oversampling factor – 1.8 1.8 1.8
Slice thickness (mm) 6 6 6 6
TR (ms) 4100 4100 3900 3900
TE (ms) 9.2 9.2 13.9 10.5
No. of repetitions 6 6 3 2
EPI factor 25 25 13 5
TSE factor – 16 16 16
Bandwidth (Hz) (frequency
direction)
3136 3160 3160 3160
SENSE factor (AP direction) 2.5 2.5 2.5 2.5
Total scan duration (minutes) 6:03 6:15 7:56 12:10
FC No Yes Yes Yes
Abbreviations: AP, anterior– posterior; TSE, turbo spin echo.

1458
|
PASCHOAL et AL.
transformed to the ASL images space, whereas for the vas-
cular tSNR analysis, we used the aBV outputs from the two-
segments 3D GRASE thresholded at 1 mL/100 g/min. The
contrast- to- noise ratio (CNR) was calculated for the perfusion
block (PLD of 1860 ms) for all readouts, according to:
where mean(GM) and mean(WM) are the mean signal over the
gray- matter and white- matter masks, and
𝜎s
is the spatial SD of
the whole- brain ASL images.
The statistical analysis was performed in R.19 A one- way
ANOVA was used to compare tSNR among different num-
bers of segments and FC levels. We also performed the two-
way ANOVA to compare CNR and blurring levels. In such
cases, the factors were the number of segments and the use or
not of deblurring. Results were adjusted for multiple compar-
isons using the Tukey’s method.
3
|
RESULTS
3.1
|
Effects of segmentation number
Two representative slices of ASL maps at multiple PLDs
are shown in Figure 2A for axial view and in Supporting
Information Figure S1 for sagittal views for the 3D- GRASE
readout with different numbers of segments and the 2D mul-
tislice scan. The sagittal maps are shown to assess blurring
in z- direction, which is especially apparent in the single-
segment GRASE scan. Figure 3 shows the tSNR maps in
axial directions for two representative PLDs that focus on the
intravascular (left) and the perfusion signal (right panel). For
(2)
CNR
=
mean (GM)−mean (WM)
𝜎
s
,
FIGURE 2 A, Two representative slices for all PLDs and readout schemes. B, Comparison of gray matter (GM) of 3D T1 images transformed
to ASL space with blurred and deblurred ASL images for each of the 3D- GRASE readouts acquired. Abbreviation: FC, flow compensation

|
1459
PASCHOAL et AL.
the PLD of 660 ms, the EPI readout had significantly higher
tSNR (0.85 ± 1.01, P < .01) than the single- segment 3D
GRASE (0.54 ± 0.60), whereas the tSNR for both the two-
segments and three- segments 3D GRASE (tSNR = 2.17 ±
3.90 and 3.10 ± 7.39, respectively), were significantly higher
than the single- segment 3D GRASE and the 2D- EPI readout
(P < .01). For the perfusion signal measured at a PLD of
1860 ms, all 3D- GRASE acquisitions showed highertSNR
than the 2D- EPI readout (tSNR = 2.94 ± 2.12, P < .01),
and the two- segments and three- segments 3D GRASE
(tSNR = 13.60 ± 12.55 and 14.88 ± 17.12, respectively)
exhibited significantly higher tSNR than single- segment 3D
GRASE (tSNR = 3.33 ± 2.33, P < .01).
The comparisons between blurred and deblurred images
for single, two, and three segments are shown in Figure 2B
(coronal view). The comparison between the two- segments
and three- segments images did not show significant differ-
ences (blurring = 42.20 ± 5.42% vs 34.86 ± 7.46%, P > .05)
for blurred images. For deblurred images, the three- segments
images showed significantly lower blurring (blurring = 5.19 ±
4.20%, P < .05) than the two- segments (11.82 ± 8.47%) and
single- segment images (10.09 ± 4.97%). The CNR measure-
ments revealed significantly higher values (P < .01) for two-
segments and three- segments 3D GRASE (CNR = 0.32 ± 0.03
and 0.33 ± 0.02, respectively), compared with single- segment
3D GRASE (0.201 ± 0.037) and 2D EPI (0.204 ± 0.023).
For a better analysis of the vascular signal, aBV maps were
calculated from the multi- PLD images for different num-
bers of segments (see Supporting Information Figure S2A).
The CBF maps were also quantified for different numbers
of segments, taking all PLDs into account (see Supporting
Information Figure S3A).
FIGURE 3 Temporal SNR (tSNR) calculated for all readouts at a PLD of 660 ms (A,C) and 1860 ms (B,D)

1460
|
PASCHOAL et AL.
3.2
|
Effects of FC
Finally, the intensity of FC gradients was varied (Figure 1D),
and the effect of the vascular signal on the visualization was
evaluated. The FC gradient scaling was performed for two-
segments 3D- GRASE te- pCASL, and the results are found
in Supporting Information Figure S4 for two representative
slices at multiple PLDs. The signal intensity maps for three
representative PLDs are shown in Figure 4. We compared the
signal intensities for different FC scales two by two. All com-
parisons were significant for a PLD of 660 ms (P < .01) ex-
cept for no FC versus FC of 25%, and FC of 75% versus FC of
100%. No significant differences were observed between FC
of 25% and FC of 50% for a PLD of 1060 ms, and between FC
of 75% and FC of 100% for a PLD of 1860 ms. Finally, aBV
and CBF maps from the respective FC scaling are shown in
Supporting Information Figures S3B and S4B, respectively.
4
|
DISCUSSION
In this study, the number of segmentations in 3D- GRASE
readout and the use of FC gradients were evaluated for
te- pCASL. Optimal settings of 3D- GRASE readout for tra-
ditional pCASL imaging for perfusion measurements (ie,
measured at a PLD of approximately 1.8 seconds) cannot be
assumed optimal for te- pCASL due to two important differ-
ences: In te- pCASL, the angiographic phase of ASL is also
captured, and the temporal footprint of te- pCASL (eight
Hadamard encodings) is longer than for traditional pCASL
(two conditions: label and control). The main findings of this
study are 2- fold. First, segmented 3D GRASE outperforms
single- segment 3D GRASE and multislice 2D EPI. Second,
the use of proper FC is advised, to allow correct visualization
of the angiographic phase.
The 2D- EPI readout showed good results for te- pCASL,
allowing the visualization of both arterial and perfusion sig-
nal (Figure 2A). However, a more careful analysis revealed
lower tSNR than segmented 3D GRASE (Figure 3). For per-
fusion signal, the tSNR of 2D EPI was lower than the one for
single- segment 3D GRASE; for the angiographic phase, 2D
EPI had a higher SNR than single- segment 3D GRASE. For
two- segmented and three- segmented 3D- GRASE images, the
tSNR values were significantly higher for both perfusion and
arterial signal when compared with the 2D EPI (Figure 2A).
When comparing the two- segments and three- segments 3D
FIGURE 4 Signal intensity calculated for different levels of FC at a PLD of 660 ms (A,D), 1060 ms (B,E), and 1860 ms (C,F). For the PLD
of 660 ms, the comparison between the FC of 25% and no FC and between FC of 50% and no FC did not show statistical difference (P < .01). For
the PLD of 1060 ms, the only comparison that did not show statistical difference (P < .01) was between the FC of 25% and FC of 50%, whereas for
the PLD of 1860 ms, the only comparison that did not result in significant difference was between FC of 75% and FC of 100%

|
1461
PASCHOAL et AL.
GRASE, the latter produced the best results. Similar results
were reported by Feinberg et al,4 who analyzed the effects of
the number of segments on single- PLD 3D- GRASE pulsed
ASL data.
The blurring effects on 3D- GRASE data were also as-
sessed. Previous studies reported a higher amount of blur-
ring when using single- segment readout for 3D- GRASE
ASL than a segmented acquisition.4,20 However, they did not
quantify the blurring, and the conclusion was based on visual
inspection. We also observed a higher amount of blurring for
single- segment 3D GRASE following visual inspection, but
a quantitative analysis did not confirm this. This is probably
because the tSNR and the spatial CNR are significantly lower
than for segmented 3D GRASE. For segmented data, two-
segments and three- segments images showed a comparable
amount of blurring before applying a deblurring algorithm.
After deblurring, the three- segments data were significantly
less blurred than the two- segments 3D- GRASE images.
The aBV maps from data with first- order FC bet-
ter delineated arterial signal. Similarly, two- segments and
three- segments GRASE provided better aBV maps than single-
segment GRASE (Supporting Information Figure S2A). As
shown in the experiment in which the flow compensation was
varied, FC use plays a big role in depicting the vascular sig-
nal.21- 24 The present results revealed that the intravascular sig-
nal visualization at PLDs of 660 ms and 1060 ms is hampered
without FC (Figure 4A,B), with FC being especially essential
at the PLD of 1060 ms (Figure 4B). At this PLD, and opposed
to the PLD of 660 ms, most of the labeled blood already filled
the large arteries within the imaging volume, while still flow-
ing fast. For the even longer PLD of 1800 ms, the blood will
have slowed down, resulting in a weaker influence of FC on
image quality, although FC of 75% and 100% still provided
slightly higher signal than lower FC levels. The smaller de-
pendency on FC when the label arrives within the microvascu-
lature and exchanges with tissue magnetization explains why
FC has not been recognized earlier as an important sequence-
design parameter. With the introduction of efficient multi- PLD
ASL sequences, this effect becomes more prominent, as these
sequences also focus on the more angiographic phase. The
fact that still higher signal was observed with FC at a PLD
of 1800 ms points to the fact that not all labels arrive at the
capillary exchange site. Inclusion of a macrovascular compo-
nent during postprocessing would correct for the label that did
not arrive yet at the voxel of their final destination, similar to
what Chappell et al has shown in a macrovascular component
that still identifies a vascular signal in vascular crushed ASL
data.25 When the label would already have arrived in the final
destination voxel, the signal should be included in quantifica-
tion to provide the best estimate of perfusion.
The main limitation of this study is the investigation of
FC gradients only in the readout direction. Future studies
may explore their application in other directions. Another
limitation is that we did not analyze the influence of motion,
as segmented 3D GRASE results in a longer temporal foot-
print, making the acquisition more prone to motion artifacts.
5
|
CONCLUSIONS
This study showed higher effective tSNR and CNR for
two- segments and three- segments GRASE compared with
2D EPI and single- segment GRASE. Therefore, its use is
recommended for te- pCASL. Moreover, FC properties of
the GRASE readout should be carefully controlled when
applying it to te- pCASL. The use of full FC is essential to
visualize the inflow of blood through the arterial system,
whereas without FC the vascular signal can be considerably
dephased, hampering the assessment of the passage of the
blood from the intravascular space to the brain tissue. With
the complete passage, data modeling can be considered op-
timal, providing more accurate CBF, aBV, and ATT maps.
When a single- segment readout is required, 3D GRASE and
multislice EPI provide similar results regarding tSNR, al-
though the tSNR for 2D EPI is higher for the angiographic
phase.
CONFLICT OF INTEREST
MJP van Osch receives research support from Philips, the
Netherlands.
ORCID
Andre M. Paschoal https://orcid.
org/0000-0001-8269-711X
Renata F. Leoni https://orcid.org/0000-0002-4568-0746
Matthias J. P. van Osch https://orcid.
org/0000-0001-7034-8959
REFERENCES
1. Alsop DC, Detre JA, Golay X, et al. Recommended implemen-
tation of arterial spin- labeled perfusion MRI for clinical applica-
tions: a consensus of the ISMRM perfusion study group and the
European consortium for ASL in dementia. Magn Reson Med.
2015;73:102- 116.
2. Oshio K, Feinberg DA. GRASE (gradient- and spin- echo) imaging:
a novel fast MRI technique. Magn Reson Med. 1991;20:344- 349.
3. Gunther M, Oshio K, Feinberg DA. Single- shot 3D imaging tech-
niques improve arterial spin labeling perfusion measurements.
Magn Reson Med. 2005;54:491- 498.
4. Feinberg D, Ramann S, Gunther M. Evaluation of new ASL 3D
GRASE sequences using parallel imaging, segmented and inter-
leaved k- space at 3T with 12- and 32- channel coils. In: Proceedings
of the 17th Annual Meeting of ISMRM, Honolulu, Hawaii, 2009.
Abstract #622.
5. Ye FQ, Frank JA, Weinberger DR, McLaughlin AC. Noise re-
duction in 3D perfusion imaging by attenuating the static signal
in arterial spin tagging (ASSIST). Magn Reson Med. 2000;44:
92- 100.

1462
|
PASCHOAL et AL.
6. Garcia DM, Duhamel G, Alsop DC. Efficiency of inversion pulses
for background suppressed arterial spin labeling. Magn Reson
Med. 2005;54:366- 372.
7. Woods JG, Chappell MA, Okell TW. A general framework for op-
timizing arterial spin labeling MRI experiments. Magn Reson Med.
2019;81:2474- 2488.
8. Gunther M, Bock M, Schad LR. Arterial spin labeling in combina-
tion with a look- locker sampling strategy: inflow turbo- sampling
EPI- FAIR (ITS- FAIR). Magn Reson Med. 2001;46:974- 984.
9. Varela M, Petersen ET, Golay X, Hajnal JV. Cerebral blood flow
measurements in infants using Look- Locker arterial spin labeling.
J Magn Reson Imaging. 2015;41:1591- 1600.
10. Gunther M. Encoded continuous arterial spin labeling. In:
Proceedings of the ISMRM Workshop on Cerebral Perfusion and
Brain Function: Novel Techniques and Applications, Salvador,
Bahia, 2007.
11. Wells JA, Lythgoe MF, Gadian DG, Ordidge RJ, Thomas DL.
In vivo Hadamard encoded continuous arterial spin labeling (H-
CASL). Magn Reson Med. 2010;63:1111- 1118.
12. Dai W, Shankaranarayanan A, Alsop DC. Volumetric measure-
ment of perfusion and arterial transit delay using Hadamard
encoded continuous arterial spin labeling. Magn Reson Med.
2013;69:1014- 1022.
13. Teeuwisse WM, Schmid S, Ghariq E, Veer IM, Van Osch MJP.
Time- encoded pseudocontinuous arterial spin labeling: basic prop-
erties and timing strategies for human applications. Magn Reson
Med. 2014;72:1712- 1722.
14. Parker DL, Goodrich KC, Roberts JA, et al. The need for
phase- encoding flow compensation in high- resolution intracra-
nial magnetic resonance angiography. J Magn Reson Imaging.
2003;18:121- 127.
15. Hennig J, Nauerth A, Friedburg H. RARE imaging: a fast imaging
method for clinical MR. Magn Reson Med. 1986;3:823- 833.
16. Chappell MA, Groves AR, Whitcher B, Woolrich MW. Variational
Bayesian inference for a nonlinear forward model. IEEE Trans
Signal Process. 2009;57:223- 236.
17. Groves AR, Chappell MA, Woolrich MW. Combined spatial and
non- spatial prior for inference on MRI time- series. Neuroimage.
2009;45:795- 809.
18. Elsayed M, Sammani F, Hamdi A, Al- Baser A, Babalghoom H. A
new method for full reference image blur measure. Int J Simul Syst
Sci Technol. 2018;19:4.
19. R Core Team. R: A Language and Environment for Statistical
Computing. Foundation for Statistical Computing. Vienna,
Austria: R Foundation for Statistical Computing; 2020. https://
www.R- proje ct.org/. Accessed March 12, 2021.
20. Vidorreta M, Wang Z, Rodriguez I, Pastor MA, Detre JA,
Fernández- Seara MA. Comparison of 2D and 3D single- shot ASL
perfusion fMRI sequences. Neuroimage. 2013;66:662- 671.
21. Wang J, Alsop DC, Song HK, et al. Arterial transit time imaging
with flow encoding arterial spin tagging (FEAST). Magn Reson
Med. 2003;50:599- 607.
22. Le TT, Fischbein NJ, André JB, Wijman C, Rosenberg J, Zaharchuk
G. Identification of venous signal on arterial spin labeling improves
diagnosis of dural arteriovenous fistulas and small arteriovenous
malformations. Am J Neuroradiol. 2012;33:61- 68.
23. Wolf RL, Wang J, Detre JA, Zager EL, Hurst RW. Arteriovenous
shunt visualization in arteriovenous malformations with arterial
spin- labeling MR imaging. Am J Neuroradiol. 2008;29:681- 687.
24. van Osch MJP, Teeuwisse WM, Chen Z, Suzuki Y, Helle M,
Schmid S. Advances in arterial spin labelling MRI methods for
measuring perfusion and collateral flow. J Cereb Blood Flow
Metab. 2018;38:1461- 1480.
25. Chappell MA, MacIntosh BJ, Donahue MJ, Günther M, Jezzard
P, Woolrich MW. Separation of macrovascular signal in multi-
inversion time arterial spin labelling MRI. Magn Reson Med.
2010;63:1357- 1365.
SUPPORTING INFORMATION
Additional Supporting Information may be found online in
the Supporting Information section.
FIGURE S1 Two representative slices for all postlabeling
delays (PLDs) and readout schemes in sagittal view
FIGURE S2 A, Arterial blood volume (aBV) maps quanti-
fied from the time- encoded pseudo- continuous arterial spin
labeling (pCASL) data in a representative subject. Color bar
shows aBV range in mL/100 g. B, The aBV maps quanti-
fied from the time- encoded pCASL data for each different
flow compensation (FC) level. Color bar shows aBV range
in mL/100 g
FIGURE S3 A, Cerebral blood flow (CBF) maps quantified
from the time- encoded pCASL data in a representative sub-
ject for the different readouts. Color bar shows CBF range in
mL/100 g/min. B, The CBF maps quantified from the time-
encoded pCASL data for each different FC level. Color bar
shows the CBF range in mL/100 g
FIGURE S4 Two representative slices for all PLDs and each
FC level acquired
How to cite this article: Paschoal AM, Leoni RF,
Pastorello BF, van Osch MJP. Three- dimensional
gradient and spin- echo readout for time- encoded
pseudo- continuous arterial spin labeling: Influence of
segmentation factor and flow compensation. Magn
Reson Med. 2021;86:1454– 1462. https://doi.
org/10.1002/mrm.28807